Correct the pitch of a vocal in your browser
This vocal pitch correction tool listens to a sung or spoken recording, finds the note you were aiming for and gently moves each one onto the key and scale you choose. Use it as a free vocal tuner online to tidy a demo take, fix a few flat notes in a cover, or push the retune speed to zero for the hard, robotic “auto tune” sound heard in pop and hip-hop.
Drop in a file or record straight from your microphone, then watch the pitch curve: the grey line is what you sang, the coloured line is the corrected result, and the faint lines are the notes of your scale. Play the before and after versions side by side and save the result as MP3, M4A or WAV. Nothing is uploaded — the analysis and the pitch shifting run on your own device.
How to use the Vocal Pitch Correction
- Drop a vocal recording on the box, press Record your voice, or press Try a demo voice to hear an out-of-tune example.
- Choose the song’s Key and Scale. Not sure? Press Detect key, or keep Chromatic, which snaps to the nearest of all twelve notes.
- Set Retune speed: 0 ms for the robotic effect, 50–150 ms for natural correction. Raise Humanize to keep your vibrato.
- Compare the Before and After players. Every change is applied again within a moment.
- Pick MP3, M4A or WAV and press Download.
Vocal Pitch Correction features
Key and scale snapping
Major, minor, harmonic minor, pentatonic, blues and modes, plus Indian raag scales such as Yaman, Bhairav, Bhairavi, Kafi and Khamaj — or chromatic for any song.
Retune speed
From an instant snap (0 ms) to a slow, transparent glide (up to 400 ms), so note starts and slides keep their character.
Humanize and strength
Keep a chosen share of your natural vibrato and expression, and pull each note only part of the way if you like.
Formants kept
The voice is re-pitched one vocal-cord period at a time, so vowels keep their natural colour instead of sounding like a chipmunk or a giant.
See the correction
A zoomable pitch curve over the scale’s note lines, before and after waveforms, and the average error in cents before and after.
Private and free
No sign-up, no watermark, no upload. Works with MP3, WAV, M4A, OGG, FLAC and WebM recordings up to 10 minutes.
How the pitch correction works
First the recording is analysed with the YIN pitch detector (de Cheveigné and Kawahara, 2002) every 5 milliseconds: it compares the waveform with delayed copies of itself and finds the delay at which it repeats, which is the period of the note. Breaths, consonants and silence have no clear period and are left untouched. Short glitches and octave jumps are smoothed out.
For every voiced moment the nearest note of your scale becomes the target. The correction moves toward it with the retune-speed time constant, and humanize corrects a 200 ms running average instead of every wiggle, so vibrato survives. The voice is then rebuilt with TD-PSOLA (pitch-synchronous overlap-add): two-period slices of the waveform are taken around each vocal-cord pulse and laid down closer together to raise the pitch, or further apart to lower it. Because each slice keeps its shape, the formants — the resonances that make an “ah” sound like an “ah” — stay where they were. Untick Keep formants and the slices are resampled instead, so the tone colour moves with the pitch.
Which settings to use
A cent is a hundredth of a semitone. Most listeners hear a sustained note as out of tune when it is about 15–25 cents away from the expected pitch, so the “notes in tune” figure counts frames within ±10 cents.
| Goal | Retune speed | Humanize | Scale |
|---|---|---|---|
| Hard robotic effect | 0 ms | 0% | The song’s key |
| Natural, invisible correction | 80–150 ms | 30–60% | The song’s key |
| Fix a few sour notes only | 40–80 ms | 20–40% | The song’s key, strength 60–80% |
| Speech or rap, light tidy-up | 100–200 ms | 50%+ | Chromatic |
| Indian classical or film song | 60–120 ms | 40%+ | Raag scale on your Sa |
Limits
- It is made for a single voice or solo instrument. A full mix with drums, chords or backing vocals confuses the pitch detector — separate the vocal first with the Vocal Extractor.
- Notes are corrected toward the nearest scale note. If you sang a different note from the one intended, it will be tuned to the wrong note; pick a narrower scale or lower the strength.
- Very breathy, growled or whispered singing has little clear pitch and is mostly left as it is.
- This page does not change timing. To move a whole song into another key, use the Audio Pitch Changer.
Vocal Pitch Correction FAQ
Is this the same as auto tune?
It does the same job as the pitch-correction plug-ins often called “auto tune”: it detects the note you sang and moves it to the nearest note of a key. It is an independent tool, not the Antares Auto-Tune product.
How do I get the robotic auto tune voice effect online for free?
Set Retune speed to 0 ms, Humanize to 0% and Strength to 100%, choose the song’s key, and download the result. The faster and more exact the snap, the more robotic it sounds.
Can I record my voice and auto tune it online?
Yes. Press Record your voice, allow the microphone, sing (ideally with headphones on), and press Stop. The take is analysed and corrected immediately; nothing is uploaded.
What key should I choose?
The key of the backing track. Press Detect key to estimate it from your singing, check it with the Key Finder, or use Chromatic, which works for any song but corrects less musically.
Will my voice sound like a chipmunk?
Not with Keep formants on. The pitch moves while the vowel resonances stay in place. Small corrections of under a semitone sound natural; large ones always sound processed.
Can I use it on a song with music?
Results are poor on full mixes because several notes sound at once. Use an a-cappella vocal stem, then mix it back with the music in your editor.
Which file should I download?
WAV if you will keep editing or mixing, MP3 or M4A for sharing. The corrected audio keeps the original sample rate and length, so it lines up with your backing track.